Why a merged PDF looks wrong, and what a merge actually keeps

A merge copies pages. It does not normalize them. If the inputs disagree about paper size or rotation, the output disagrees too. Viewers hide that by scaling every page to the window. Printers, upload portals, and page-numbering tools do not.

What a merge keeps, measured

We merged three files with Merge PDF: a 2-page US Letter document, a 2-page A4 document, and a 1-page phone-style scan saved as a landscape page (792×612 pt) with a /Rotate 90 flag. Then we opened the 5-page result with pikepdf and checked every page:

• Pages 1 and 2 kept their Letter size, 612×792 pt.
• Pages 3 and 4 kept their A4 size, 595.28×841.89 pt.
• Page 5 kept its landscape box and its /Rotate 90 flag.

Nothing was resized, cropped, or re-rendered. Text stays selectable and images keep their resolution. That is correct behavior for a merge, and it is also why a merged packet can print badly.

Page size mismatches

US Letter is 612×792 pt (8.5×11 in). A4 is 595.28×841.89 pt (210×297 mm). Letter is 16.7 pt wider and 49.9 pt shorter. On screen both look full. On paper:

• Printed at actual size on Letter stock, an A4 page loses about 17.6 mm of height. If the printer centers the page, that is split between the top and bottom. Either way, a footer or page number can be cut off.
• Printed with “fit to page”, the A4 page is scaled to 94.1% and picks up white strips at the sides.

Neither is a merge bug. Fix it before combining: re-export the odd pages at the paper size the recipient will print on. If the source is photos, create the PDF with a fixed paper size in JPG to PDF rather than Fit. The Letter vs A4 guide has the numbers in both directions.

Orientation conflicts

Phone scanner apps often save a page as a landscape box plus a /Rotate 90 instruction instead of a portrait page. Chrome, Acrobat, and most viewers apply the rotation, so you see an upright page. Some print drivers, preview panes, and document-management systems handle /Rotate inconsistently and show the page sideways. A merge keeps the flag exactly as it was, as the test above shows, so the problem travels into the merged file.

Check the thumbnails in the merge list before you download. If a page looks sideways there, fix that one file first (re-export it upright from the scanner app) and add it again.

What a merge drops, measured

Page content survives. Document-level structure does not. We merged a 3-page file that had a title (“Contract A”), an author, an XMP metadata block, three bookmarks, and an internal link on page 1 pointing to page 3, together with a 2-page Letter file. In the merged output:

• Bookmarks were gone. The output had no outline at all.
• Title, author, subject, keywords were gone. The document properties showed Creator and Producer as “UnlimitedPDF” and the creation and modification dates.
• XMP metadata was gone.
• The internal link was still on page 1 and still clickable, but its destination pointed at a copy of page 3 that is not part of the merged page list. In practice it goes nowhere useful, and that orphaned copy also adds a little to the file size.

The metadata part is often good news: a merged packet does not carry “Author: mchen” or an old draft title from one of its inputs. If you want a title in the file, set it afterwards in a PDF editor. If you need a table of contents or working cross-references, rebuild them in the merged file with an editor that supports bookmarks.

Earlier saved versions inside a source file (incremental updates after an old %%EOF marker) are not carried over either, because only pages from the current page tree are copied. PDF X-Ray can confirm that on the output.

Blank and duplicate pages

Sheet feeders produce blank pages, and a blank page from a color scan is not empty: it is a full-page image of white paper, often with visible noise and a real file size. Expand the file in the merger and click a page thumbnail to exclude it. The same works for a duplicated cover page or a separator sheet.

File size after a merge

A merge does not compress anything. The output is roughly the sum of the inputs, minus the pages you excluded, plus a little structure. If the packet has to fit under an email or portal limit, compress the merged file once at the end rather than each input separately, and check the result against the limit. The email guide covers the arithmetic.

Checklist before merging

1. Decide the paper size the recipient will print on, and make every input match it.
2. Open each scan in a viewer and re-export any page that shows up sideways.
3. Put the files in order, then expand them to drop blank or duplicate pages.
4. Merge, then scroll through the output once at page-width zoom.
5. Compress the merged file if it has to meet a size limit.
6. Add bookmarks or a document title afterwards if the recipient needs them.

How we measured

The merge tests were run on 27 September 2026 with this site’s Merge PDF tool in Chrome on a Linux desktop. Test files were generated with ReportLab and pikepdf. Page sizes, rotation, document properties, XMP, outlines, and link destinations in the outputs were read with pikepdf.